(1) Field of the Invention
The present invention relates to a speech recognition dictionary creation apparatus for creating a speech recognition dictionary that enables an information search apparatus such as a program guidance apparatus to perform speech recognition and the information search apparatus.
(2) Description of the Related Art
Recently, a conventional program guidance apparatus which includes a speech recognition dictionary creation apparatus has come in use with the increase in the number of program channels, and the conventional program guidance apparatus is for operating TV using an Electronic Program guide (EPG) and speech, receiving latest EPG data, creating a speech recognition dictionary corresponding to a program to be broadcast today or on a later date, and then deleting an old recognition dictionary so as not to use any extra hard resources.
Further, at the time of creating a dictionary, it classifies each of words into a certain category such as “program name”, “genre”, and “cast” in the EPG so as to register them in the dictionary and recognizes each of the words equally.
FIG. 1 is a block diagram showing the structure of the program guidance apparatus that has a conventional speech recognition dictionary creation apparatus.
The program guidance apparatus comprises an EPG receiving unit 701, an item classifying unit 702, a dictionary updating unit 703, a dictionary 704, a speech recognition unit 705, a clock 706, a registered word deleting unit 707, an EPG storage unit 708 and a control unit 709.
The EPG receiving unit 701 demultiplexes EPG data from a TV signal. The item classifying unit 702 demultiplexes EPG data such as “broadcasting time” that does not need for speech recognition and classifies the residual EPG data into a category such as “program title” and “cast”. The dictionary updating unit 703 registers words included in the EPG data in a dictionary without duplication. The dictionary 704 records the words necessary for speech recognition along with its registration date and time. The speech recognition unit 705 refers to the dictionary 704 based on a speech signal by user speech and recognizes the specific words.
Also, the clock 706 outputs the present date and time to the registered word deleting unit 707. The registered word deleting unit 707 refers to the date and time of the clock 706 and deletes, for example, the words registered up to the previous day in all the words registered in the dictionary 704. The EPG storage unit 708 stores the EPG data demultiplexed in the EPG receiving unit 701. The control unit 709 searches the EPG data stored in the EPG storage unit 708 for the words recognized by the speech recognition unit 705 and outputs the corresponding program contents (for example, refer to FIG. 1 in page 7 of the Japanese Laid-Open Patent application No. 2001-22374).
The contents of recognition words in a dictionary changes because the conventional program guidance apparatus mentioned above creates a dictionary using latest EPG data, but its user rarely pays attention to the change.
Here, for example, in the case where a user performs a TV program search or TV operation using a program name that has been viewed by the user and has interested the user or a cast name that is not included in the latest EPG data, the program guidance apparatus outputs a word that has the highest recognition score in the dictionary as the recognition result. This is because any word that is not registered in the dictionary cannot be recognized. Therefore, the outputted recognition result may differ from the program name or the cast name that is expected by the user. In this case, the user is to think that this recognition result is an error and say the same words repeatedly.
Therefore, it is conceivable to use not only the latest EPG data but also past EPG data in creating a dictionary, but in this case, a large number of words must be registered in the dictionary.
Also, with the increase in the number of program channels, the number of the recognition words also increases. However, there is a problem that its recognition rate is decreasing because the conventional speech recognition dictionary creation apparatus searches only new data as words to be recognized.